Since the Human Microbiome Project1 launched in 2008, we’ve learned from hundreds of studies that the microbiome — all of the bacteria, archaea, viruses and fungi that live on us and in us — is associated with human disease. Inflammatory bowel disease (IBD) and obesity are some examples of interest to gastroenterologists where the microbiome clearly has an impact. As a result, there is enthusiasm for leveraging the therapeutic potential of the microbiota. However, we still have much to learn before the microbiome becomes a treatment option for clinicians. Statistical power is key, and therefore, so are large sample sizes.
The American Gut Project2, the country’s largest open source citizen science project, is making great strides toward this goal. As of May 2016, the project has processed over 9,000 samples from over 7,000 individuals. To continue growing and diversifying the project, American Gut has reached out to various groups along the health spectrum such as patients, athletes, and now physicians thanks to a collaboration with the AGA and its Center for Gut Microbiome Research and Education3.
At Digestive Disease Week® (DDW) 2016, the center hosted a one-of-a-kind session titled Active Learning Session on the Gut Microbiome: Fundamental Principles in Theory and Practice. AGA member volunteers submitted stool samples to American Gut for analysis in advance of the session. In addition to the individual results that participants would normally receive, American Gut also performed a special cohort analysis to see if the guts of gastroenterologists differ from those of our larger volunteer pool.
The American Gut dataset, including AGA volunteer samples, was compared with four other human datasets: a U.S. IBD cohort including both Crohn’s and ulcerative colitis patients5, an adolescent Crohn’s cohort7, a Chinese Crohn’s disease cohort4, and a single individual diagnosed with Crohn’s disease with longitudinal samples spanning months6. We also included a canine IBD dataset to assess species-specific differences and similarities. Leveraging Qiita, an open-source microbiome data storage and analysis resource developed in our lab8, we conducted a meta-analysis of all of these studies together.
Figure 1 shows a principal coordinates analysis (PCoA) plot of the meta-analysis. Each dot on the plot represents a single sample. The closer together two samples are on the plot, the more similar the microbial communities represented by those dots; the further apart two samples are, the more dissimilar the microbial communities.
The first thing you may notice is that the AGA samples (red) are widely distributed, as are the American Gut samples (dark blue, small dots). Most AGA samples came from individuals that do not suffer from IBD. Notably, the longitudinal samples from the individual Crohn’s patient (yellow) clustered separately from the U.S. IBD cohort (light blue), the adolescent Crohn’s cohort (purple), and the Chinese Crohn’s cohort (green).
The Chinese cohort also forms an obvious cluster; this is likely due to differences in diet between individuals in the Chinese cohort and the other cohorts, which were largely North American patients with “Western” diets. Interestingly, some samples from the U.S. and adolescent IBD cohorts clustered with the canine samples, indicating that microbially these samples were more similar to dog samples than to the other human samples in the meta-analysis.
The datasets can also be visualized differently to look for patterns of clustering across studies. Here, additional visualization by IBD diagnosis (Figure 2) and by IBD subtype (Figure 3) are shown. Neither revealed obvious clustering patterns, indicating that overall microbiome composition is not notably associated with specific IBD subtypes among the seven studies in this meta-analysis.
However, this may be due to the large size of the American Gut cohort (over 6,000 fecal samples) compared to the other studies. Additionally, any specific associations by IBD subtype could be present at levels (e.g., at the individual species or strain level) not detectable with the methodology used in this analysis. It is important to note that American Gut participants self-report diseases and their reported diagnoses are not independently verified. This underscores the potential benefits of collecting samples from well-characterized diseased cohorts to fully piece together the connection between the microbiome and IBD.
This is just a glimpse into the findings from our collaboration with AGA and additional insights are available on http://www.gastro.org/about/initiatives/AGA-American_Gut_ Handout_DDW_2016.pdf.9 A recording of the center’s DDW 2016 session is also available to view at DDW On Demand. At American Gut, we look forward to continued analyses and building connections between the human microbiome and health. We will continue to share our knowledge and findings with AGA members through the AGA Center for Gut Microbiome Research and Education.
Dr. Knight has no conflicts to disclose.
4. Courtesy of Hongwei Zhou and Ye Chen
5. Courtesy of William Sandborn
6. Courtesy of Larry Smarr
7. D. Gevers, S. Kugathasan, L. A. Denson, Y. Vazquez-Baeza, W. Van Treuren, B. Ren, E. Schwager, D. Knights, S. J. Song, M. Yassour, X. C. Morgan, A. D. Kostic, C. Luo, A. Gonzalez, D. McDonald, Y. Haberman, T. Walters, S. Baker, J. Rosh, M. Stephens, M. Heyman, J. Markowitz, R. Baldassano, A. Griffiths, F. Sylvester, D. Mack, S. Kim, W. Crandall, J. Hyams, C. Huttenhower, R. Knight and R. J. Xavier, “The treatment-naive microbiome in new-onset Crohn’s disease,” Cell Host Microbe, vol. 15, no. 3, pp. 382-392, 2014.
8. Qiita’s main server is accessible at http://qiita. microbio.me
9 http://www.gastro.org/about/initiatives/AGA-American_ Gut_Handout_DDW_2016.pdf